Surfacing Insights: Balancing Signal and Noise
I explore how agentic AI works behind the scenes to combine signals across a GRC platform, suggesting next-best actions that go beyond what a standalone risk view can provide.
At a glance
• Explores how agentic AI can synthesize signals across the platform, not just within a single risk tool
• Shifts interpretation and prioritization from the user to the system, suggesting clearer next-best actions
Why this experiment
Enterprise risk tools are often stable, trusted, and widely adopted—but they also tend to calcify. Over time, successful patterns become fixed views that are hard to evolve, even as the volume, velocity, and interconnectedness of risk data increases.
This experiment starts from one of those stable patterns: a high-density risk heatmap designed to summarize exposure at a glance. Rather than replacing it, the goal was to explore how that familiar surface could be extended—making it more scannable, more informative, and more adaptive—without breaking user trust or introducing unnecessary complexity.
The experiment also creates space to explore a broader question: how a platform-level, agentic AI could operate behind the scenes to add context and guidance, instead of simply reflecting what a single risk tool already knows.
The core problem
Traditional risk views are largely one-to-one mappings: a risk object produces a corresponding set of insights, typically confined to the boundaries of the risk module itself. While this works for visibility, it breaks down when users need to understand: • how risks relate to activity elsewhere in the platform • which signals actually matter right now • what action, if any, should be taken next
As risk environments grow more complex, these views ask users to do too much cognitive work—scanning dense layouts, interpreting raw indicators, and manually connecting signals across tools.
The core problem, then, is not a lack of data, but a lack of contextual prioritization and guidance: how to surface the right signals at the right moment, and how to suggest meaningful next steps without overwhelming the user.
This experiment explores what happens when that responsibility shifts—from the user assembling context manually, to an agentic system synthesizing signals across the platform and quietly shaping what is shown, emphasized, or suggested.